
#importing all the necessary mathematical, data handling & visulaization libraries
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
sns.set_theme(style="whitegrid")
import scipy.stats as stats
from sklearn.preprocessing import MinMaxScaler, StandardScaler
pd.set_option('display.max_rows', None)
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix
from scipy.stats import zscore
from collections import Counter


df_tel1 = pd.read_csv('TelcomCustomer-Churn_1.csv')                      #Answer


df_tel1.head()


df_tel1.shape

(7043, 10)


df_tel2 = pd.read_csv('TelcomCustomer-Churn_2.csv')                              #Answer


df_tel2.shape

(7043, 12)


df1_tel_merged = pd.merge(df_tel1, df_tel2, how= 'outer', on= 'customerID' )     #Answer


df1_tel_merged.shape

(7043, 21)


df1_tel_merged.head(2)


df1_tel_merged.dtypes

customerID           object
gender               object
SeniorCitizen         int64
Partner              object
Dependents           object
tenure                int64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges         object
Churn                object
dtype: object


df1_tel_merged.columns == df1_tel_merged.columns                       #Answer

array([ True,  True,  True,  True,  True,  True,  True,  True,  True,
        True,  True,  True,  True,  True,  True,  True,  True,  True,
        True,  True,  True])


df1_tel_merged == df1_tel_merged.columns

C:\Users\user\Anaconda3\lib\site-packages\pandas\core\ops\array_ops.py:253: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
  res_values = method(rvalues)


df1_tel_merged.isnull().sum()

customerID          0
gender              0
SeniorCitizen       0
Partner             0
Dependents          0
tenure              0
PhoneService        0
MultipleLines       0
InternetService     0
OnlineSecurity      0
OnlineBackup        0
DeviceProtection    0
TechSupport         0
StreamingTV         0
StreamingMovies     0
Contract            0
PaperlessBilling    0
PaymentMethod       0
MonthlyCharges      0
TotalCharges        0
Churn               0
dtype: int64


df1_tel_merged.empty

False


df1_tel_merged.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]


print('There are no null values in the dataframe, but we can see there are empty strings in the TotalCharges column in these locations 488,753,936,1082,1340,3331,3826,4380,5218,6670,6754 ')

There are no null values in the dataframe, but we can see there are empty strings in the TotalCharges column in these locations 488,753,936,1082,1340,3331,3826,4380,5218,6670,6754


#df1_tel_merged.replace('', np.nan)


#df1_tel_merged['TotalCharges'].mean()


df1_tel_merged = df1_tel_merged.replace(r'^\s*$', np.nan, regex= True) #Answer - Replacing the empty strings values with NaN values


df1_tel_merged.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]


df1_tel_merged.isna().sum()  # We replaced null values in TotalCharges column wherever empty strings were there. Total there are 11 records

customerID           0
gender               0
SeniorCitizen        0
Partner              0
Dependents           0
tenure               0
PhoneService         0
MultipleLines        0
InternetService      0
OnlineSecurity       0
OnlineBackup         0
DeviceProtection     0
TechSupport          0
StreamingTV          0
StreamingMovies      0
Contract             0
PaperlessBilling     0
PaymentMethod        0
MonthlyCharges       0
TotalCharges        11
Churn                0
dtype: int64


df1_tel_merged['TotalCharges'].replace(np.nan, 2283, inplace= True) #Answer - Imputing average value of 2283 NaN values


df1_tel_merged.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]


df1_tel_merged.describe().T


df1_tel_merged.head(2)


df1_tel_merged._get_numeric_data().columns                       #Also, total charges is also a continuous variable as it has lot of float values

Index(['SeniorCitizen', 'tenure', 'MonthlyCharges'], dtype='object')


df1_tel_merged = df1_tel_merged.astype({'tenure': 'float64',         #Answer
                                        'SeniorCitizen': 'float64',
                                        'MonthlyCharges': 'float64',
                                        'TotalCharges' : 'float64'})


df1_tel_merged.dtypes

customerID           object
gender               object
SeniorCitizen       float64
Partner              object
Dependents           object
tenure              float64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges        float64
Churn                object
dtype: object


df1_tel_merged.head(2)


print(Counter(df1_tel_merged['MultipleLines']))

Counter({'No': 3390, 'Yes': 2971, 'No phone service': 682})


import matplotlib.ticker as ticker
import matplotlib.cm as cm
import matplotlib as mpl
from matplotlib.gridspec import GridSpec
import matplotlib.pyplot as plt


#StreamingTV_type = telecomcustomer_merged_df.groupby('StreamingTV').agg('count')
#MultipleLines = telecomcustomer_merged_df.groupby('MultipleLines').agg('count')
#PaymentMethod = telecomcustomer_merged_df.groupby('PaymentMethod').agg('count')
#InternetService = telecomcustomer_merged_df.groupby('InternetService').agg('count')
#Dependents = telecomcustomer_merged_df.groupby('Dependents').agg('count')
#StreamingMovies = telecomcustomer_merged_df.groupby('StreamingMovies').agg('count')
#Contract = telecomcustomer_merged_df.groupby('Contract').agg('count')
#OnlineBackup = telecomcustomer_merged_df.groupby('OnlineBackup').agg('count')
#DeviceProtection = telecomcustomer_merged_df.groupby('DeviceProtection').agg('count')
#TechSupport = telecomcustomer_merged_df.groupby('TechSupport').agg('count')
#Partner = telecomcustomer_merged_df.groupby('Partner').agg('count')
#PaperlessBilling = telecomcustomer_merged_df.groupby('PaperlessBilling').agg('count')
#PhoneService = telecomcustomer_merged_df.groupby('PhoneService').agg('count')
#Churn = telecomcustomer_merged_df.groupby('Churn').agg('count')
#gender = telecomcustomer_merged_df.groupby('gender').agg('count')
#OnlineSecurity = telecomcustomer_merged_df.groupby('OnlineSecurity').agg('count')


def telecomcustomer_merged_df_pie_chart_function(x):                       #Answer
    num_cols = x._get_numeric_data().columns 
    num_cols = set(num_cols)
    normal_cols = set(x.columns)
    cat_cols = list(normal_cols - num_cols)
    cat_cols.remove('customerID')
    print(cat_cols)
    
    
    for feature in cat_cols[0:16]:
        feature_type = df1_tel_merged.groupby(feature).agg('count')
        feature_labels = feature_type.customerID.sort_values().index
        feature_counts = feature_type.customerID.sort_values()
        print('\n')
        print(feature_labels)
        print('\n')
        print(feature_counts)
        
        plt.figure(1, figsize=(40,20)) 
        the_grid = GridSpec(2, 2)
        cmap = plt.get_cmap('Accent')
        colors = [cmap(i) for i in np.linspace(0, 1, 8)]
        
        title_categorical_column_set = {'StreamingTV'      : 'Types of StreamingTV',
                                        'MultipleLines'    : 'Types of MultipleLines',
                                        'PaymentMethod'    : 'Types of payment method',
                                        'InternetService'  : 'Types of InternetService',
                                        'Dependents'       : 'Types of Dependents',
                                        'StreamingMovies'  : 'Types of StreamingMovies',
                                        'Contract'         : 'Types of Contracts',
                                        'OnlineBackup'     : 'Types of OnlineBackup',
                                        'DeviceProtection' : 'Types of DeviceProtection',
                                        'TechSupport'      : 'Types of TechSupport',
                                        'Partner'          : 'Partner',
                                        'PaperlessBilling' : 'Types of PaperlessBilling',
                                        'PhoneService'     : 'Types of PhoneService',
                                        'Churn'            : 'Types of churn',
                                        'gender'           : 'Types of gender',
                                        'OnlineSecurity'   : 'Types of OnlineSecurity'}
                
        plt.subplot(the_grid[0, 1], aspect=1, title= title_categorical_column_set[feature])
        type_feature = plt.pie(feature_counts, labels= feature_labels, autopct='%1.1f%%', shadow=True, colors=colors)
        plt.show()  
        print('*************************************************************************************************************')


telecomcustomer_merged_df_pie_chart_function(df1_tel_merged)

['Dependents', 'Contract', 'PaperlessBilling', 'OnlineBackup', 'InternetService', 'StreamingMovies', 'OnlineSecurity', 'PhoneService', 'TechSupport', 'gender', 'StreamingTV', 'MultipleLines', 'DeviceProtection', 'Churn', 'PaymentMethod', 'Partner']


Index(['Yes', 'No'], dtype='object', name='Dependents')


Dependents
Yes    2110
No     4933
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['One year', 'Two year', 'Month-to-month'], dtype='object', name='Contract')


Contract
One year          1473
Two year          1695
Month-to-month    3875
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No', 'Yes'], dtype='object', name='PaperlessBilling')


PaperlessBilling
No     2872
Yes    4171
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No internet service', 'Yes', 'No'], dtype='object', name='OnlineBackup')


OnlineBackup
No internet service    1526
Yes                    2429
No                     3088
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No', 'DSL', 'Fiber optic'], dtype='object', name='InternetService')


InternetService
No             1526
DSL            2421
Fiber optic    3096
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No internet service', 'Yes', 'No'], dtype='object', name='StreamingMovies')


StreamingMovies
No internet service    1526
Yes                    2732
No                     2785
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No internet service', 'Yes', 'No'], dtype='object', name='OnlineSecurity')


OnlineSecurity
No internet service    1526
Yes                    2019
No                     3498
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No', 'Yes'], dtype='object', name='PhoneService')


PhoneService
No      682
Yes    6361
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['No internet service', 'Yes', 'No'], dtype='object', name='TechSupport')


TechSupport
No internet service    1526
Yes                    2044
No                     3473
Name: customerID, dtype: int64

*************************************************************************************************************


Index(['Female', 'Male'], dtype='object', name='gender')


gender
Female    3488
Male      3555
Name: customerID, dtype: int64


#Answer
print('1. For StreamingTV, more than 50 % customers dont have streaming service nor they have internet connection')
print('\n')
print('2. For Multiple Lines, almost 58 % customers neither have phone service nor have any multiple lines connections')
print('\n')
print('3. For Payment method, multiple customers use different ways to pay, but majority of them use electronic check clearence')
print('\n')
print('4. For Internet Service, around  78 % customers use internet service where Fibre Optic is most used having 34%')
print('\n')
print('5. For Dependents, around  70 % customers are non dependents and are single operated. 30 % use sharing option and have dependents')
print('\n')
print('6. For Streaming Movies, almost 61 % customers neither have internet service nor have streaming movies subscription')
print('\n')
print('7. For Contract, around 55% customers use monthly plans/contracts and 45% have long term plans like 1 year/2years')
print('\n')
print('8. For Online Backup, around 35% customers only have online backups')
print('\n')
print('9. For Device protection, around 35% customers only have opted for device protection and rest neither have internet service nor have device protection ') 
print('\n')
print('10. For Tech support, only 29% customer have opted for tech support ')
print('\n')
print('11. For Partner, there is nearly 50 - 50 % proportion')
print('\n')
print('12. For Paperless Billing, 60% customer opted for paperless bills')
print('\n')
print('13. For Phone Service, mojority of people have phone service enabled - 90%')
print('\n')
print('14. For Churn, almost 75% people have not left their existing service but 25% have left the service')
print('\n')
print('15. For gender, there is nearly 50 - 50 % proportion between Males and Females')
print('\n')
print('16. For Online Security, 70% neither have internet service nor have opted for the security service')

1. For StreamingTV, more than 50 % customers dont have streaming service nor they have internet connection


2. For Multiple Lines, almost 58 % customers neither have phone service nor have any multiple lines connections


3. For Payment method, multiple customers use different ways to pay, but majority of them use electronic check clearence


4. For Internet Service, around  78 % customers use internet service where Fibre Optic is most used having 34%


5. For Dependents, around  70 % customers are non dependents and are single operated. 30 % use sharing option and have dependents


6. For Streaming Movies, almost 61 % customers neither have internet service nor have streaming movies subscription


7. For Contract, around 55% customers use monthly plans/contracts and 45% have long term plans like 1 year/2years


8. For Online Backup, around 35% customers only have online backups


9. For Device protection, around 35% customers only have opted for device protection and rest neither have internet service nor have device protection 


10. For Tech support, only 29% customer have opted for tech support 


11. For Partner, there is nearly 50 - 50 % proportion


12. For Paperless Billing, 60% customer opted for paperless bills


13. For Phone Service, mojority of people have phone service enabled - 90%


14. For Churn, almost 75% people have not left their existing service but 25% have left the service


15. For gender, there is nearly 50 - 50 % proportion between Males and Females


16. For Online Security, 70% neither have internet service nor have opted for the security service


replace_structure = {'gender'          : {'Male' : 0, 'Female' : 1},
                     'Partner'         : {'Yes' : 1, 'No' : 0},
                     'Dependents'      : {'Yes' : 1, 'No' : 0},
                     'PhoneService'    : {'Yes' : 1, 'No' :0},
                     'MultipleLines'   : {'No phone service' : -1, 'Yes' : 1, 'No' : 0},
                     'InternetService' : {'No' : 0, 'DSL' : 1, 'Fiber optic' : 2},
                     'OnlineSecurity'  : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'DeviceProtection' : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'TechSupport'     : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'StreamingTV'     : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'StreamingMovies' : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'Contract'        : {'Month-to-month' : 0, 'One year' : 1, 'Two year' : 2},
                     'OnlineBackup'    : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'PaperlessBilling': {'Yes' : 1, 'No' :0},
                     'PaymentMethod'   : {'Electronic check' : 0, 'Mailed check' : 1, 'Bank transfer (automatic)' : 2, 'Credit card (automatic)' : 3},
                     'Churn'           : {'Yes' : 1, 'No' :0} }


df1_tel_merged = df1_tel_merged.replace(replace_structure)           #Answer - All the values are encoded with single digits


df1_tel_merged.head()


X = df1_tel_merged.drop(labels= ['Churn','customerID'], axis=1)       
y = df1_tel_merged['Churn']


X.head()


y.head()

0    0
1    0
2    1
3    0
4    1
Name: Churn, dtype: int64


X_train, X_test, y_train, y_test = train_test_split(X, y, test_size= 0.20, random_state= 1 )    #Answer


print(X_train.shape)
print(X_test.shape)
print(y_train.shape)
print(y_test.shape)

(5634, 19)
(1409, 19)
(5634,)
(1409,)


# However, random forest/ensemble methods methods won't have impact with outliers, we will still convert the  features data X as a standard approach, as we have some nummerical features like Total Charges which have high values
X_train_scaled = X_train.apply(zscore)              #Answer
X_test_scaled = X_test.apply(zscore)


X_train_scaled.head()


X_test_scaled.head()


import xgboost as xgb  #import xgbost model


xgboost_model_1 = xgb.XGBClassifier()          # Instantiating the model


xgboost_model_1.fit(X_train_scaled, y_train)               # Answer - Fiting/traing the model

C:\Users\user\Anaconda3\lib\site-packages\xgboost\sklearn.py:1146: UserWarning: The use of label encoder in XGBClassifier is deprecated and will be removed in a future release. To remove this warning, do the following: 1) Pass option use_label_encoder=False when constructing XGBClassifier object; and 2) Encode your labels (y) as integers starting with 0, i.e. 0, 1, 2, ..., [num_class - 1].
  warnings.warn(label_encoder_deprecation_msg, UserWarning)

[01:32:29] WARNING: D:\bld\xgboost-split_1631904903843\work\src\learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'binary:logistic' was changed from 'error' to 'logloss'. Explicitly set eval_metric if you'd like to restore the old behavior.

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=1, gamma=0, gpu_id=-1,
              importance_type='gain', interaction_constraints='',
              learning_rate=0.300000012, max_delta_step=0, max_depth=6,
              min_child_weight=1, missing=nan, monotone_constraints='()',
              n_estimators=100, n_jobs=4, num_parallel_tree=1, random_state=0,
              reg_alpha=0, reg_lambda=1, scale_pos_weight=1, subsample=1,
              tree_method='exact', validate_parameters=1, verbosity=None)


predictions_1 = xgboost_model_1.predict(X_test_scaled)                    # Predicting th results


print(confusion_matrix(y_test, predictions_1))                                                #Answer
print('\n')
print(classification_report(y_test, predictions_1))

[[923 138]
 [159 189]]


              precision    recall  f1-score   support

           0       0.85      0.87      0.86      1061
           1       0.58      0.54      0.56       348

    accuracy                           0.79      1409
   macro avg       0.72      0.71      0.71      1409
weighted avg       0.79      0.79      0.79      1409


print('We got 79% Accuracy with weighted average 79% in Precision, 79% in recall and 79% in F1-score. However, we can see some imbalance in the classes where 1 has very less data ')

We got 79% Accuracy with weighted average 79% in Precision, 79% in recall and 79% in F1-score. However, we can see some imbalance in the classes where 1 has very less data


print(xgboost_model_1.score(X_train_scaled, y_train))
print(xgboost_model_1.score(X_test_scaled, y_test))

0.9309549165779197
0.7892122072391767


from sklearn.model_selection import GridSearchCV


xgboost_model_tuning = xgb.XGBClassifier()


parameters = {
     "eta"    : [0.05, 0.10, 0.15, 0.20, 0.25 ] ,
     "max_depth"        : [ 3, 4, 5, 6],
     "min_child_weight" : [ 1, 3, 5 ],
     "gamma"            : [ 0.0, 0.1],
     "colsample_bytree" : [ 0.3, 0.4],
     "booster"          : ['gbtree']
           }


 xgboost_best_model = GridSearchCV(xgboost_model_tuning,
                                    parameters,
                                    n_jobs=3,
                                    cv=4)


xgboost_best_model.fit(X_train_scaled, y_train)

C:\Users\user\Anaconda3\lib\site-packages\xgboost\sklearn.py:1146: UserWarning: The use of label encoder in XGBClassifier is deprecated and will be removed in a future release. To remove this warning, do the following: 1) Pass option use_label_encoder=False when constructing XGBClassifier object; and 2) Encode your labels (y) as integers starting with 0, i.e. 0, 1, 2, ..., [num_class - 1].
  warnings.warn(label_encoder_deprecation_msg, UserWarning)

[01:36:01] WARNING: D:\bld\xgboost-split_1631904903843\work\src\learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'binary:logistic' was changed from 'error' to 'logloss'. Explicitly set eval_metric if you'd like to restore the old behavior.

GridSearchCV(cv=4,
             estimator=XGBClassifier(base_score=None, booster=None,
                                     colsample_bylevel=None,
                                     colsample_bynode=None,
                                     colsample_bytree=None, gamma=None,
                                     gpu_id=None, importance_type='gain',
                                     interaction_constraints=None,
                                     learning_rate=None, max_delta_step=None,
                                     max_depth=None, min_child_weight=None,
                                     missing=nan, monotone_constraints=None,
                                     n_estimators=100, n_jobs=None,
                                     num_parallel_tree=None, random_state=None,
                                     reg_alpha=None, reg_lambda=None,
                                     scale_pos_weight=None, subsample=None,
                                     tree_method=None, validate_parameters=None,
                                     verbosity=None),
             n_jobs=3,
             param_grid={'booster': ['gbtree'], 'colsample_bytree': [0.3, 0.4],
                         'eta': [0.05, 0.1, 0.15, 0.2, 0.25],
                         'gamma': [0.0, 0.1], 'max_depth': [3, 4, 5, 6],
                         'min_child_weight': [1, 3, 5]})


df_xgboost_best_model_results = pd.DataFrame(xgboost_best_model.cv_results_)


df_xgboost_best_model_results.head()


df_xgboost_best_model_results[['param_colsample_bytree','param_eta','param_gamma','param_max_depth', 'param_min_child_weight','params','mean_test_score']]


print(xgboost_best_model.best_estimator_)               #Answer
print('\n')
print(xgboost_best_model.get_params)
print('\n')
print(xgboost_best_model.best_score_)

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=0.3, eta=0.1, gamma=0.1,
              gpu_id=-1, importance_type='gain', interaction_constraints='',
              learning_rate=0.100000001, max_delta_step=0, max_depth=3,
              min_child_weight=5, missing=nan, monotone_constraints='()',
              n_estimators=100, n_jobs=4, num_parallel_tree=1, random_state=0,
              reg_alpha=0, reg_lambda=1, scale_pos_weight=1, subsample=1,
              tree_method='exact', validate_parameters=1, verbosity=None)


<bound method BaseEstimator.get_params of GridSearchCV(cv=4,
             estimator=XGBClassifier(base_score=None, booster=None,
                                     colsample_bylevel=None,
                                     colsample_bynode=None,
                                     colsample_bytree=None, gamma=None,
                                     gpu_id=None, importance_type='gain',
                                     interaction_constraints=None,
                                     learning_rate=None, max_delta_step=None,
                                     max_depth=None, min_child_weight=None,
                                     missing=nan, monotone_constraints=None,
                                     n_estimators=100, n_jobs=None,
                                     num_parallel_tree=None, random_state=None,
                                     reg_alpha=None, reg_lambda=None,
                                     scale_pos_weight=None, subsample=None,
                                     tree_method=None, validate_parameters=None,
                                     verbosity=None),
             n_jobs=3,
             param_grid={'booster': ['gbtree'], 'colsample_bytree': [0.3, 0.4],
                         'eta': [0.05, 0.1, 0.15, 0.2, 0.25],
                         'gamma': [0.0, 0.1], 'max_depth': [3, 4, 5, 6],
                         'min_child_weight': [1, 3, 5]})>


0.8042253986144268


xgboost_final_model = xgb.XGBClassifier(colsample_bytree=0.4, eta=0.1, gamma=0.1, max_depth=4, min_child_weight=5)


xgboost_final_model.fit(X_train_scaled, y_train)

C:\Users\user\Anaconda3\lib\site-packages\xgboost\sklearn.py:1146: UserWarning: The use of label encoder in XGBClassifier is deprecated and will be removed in a future release. To remove this warning, do the following: 1) Pass option use_label_encoder=False when constructing XGBClassifier object; and 2) Encode your labels (y) as integers starting with 0, i.e. 0, 1, 2, ..., [num_class - 1].
  warnings.warn(label_encoder_deprecation_msg, UserWarning)

[01:36:04] WARNING: D:\bld\xgboost-split_1631904903843\work\src\learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'binary:logistic' was changed from 'error' to 'logloss'. Explicitly set eval_metric if you'd like to restore the old behavior.

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=0.4, eta=0.1, gamma=0.1,
              gpu_id=-1, importance_type='gain', interaction_constraints='',
              learning_rate=0.100000001, max_delta_step=0, max_depth=4,
              min_child_weight=5, missing=nan, monotone_constraints='()',
              n_estimators=100, n_jobs=4, num_parallel_tree=1, random_state=0,
              reg_alpha=0, reg_lambda=1, scale_pos_weight=1, subsample=1,
              tree_method='exact', validate_parameters=1, verbosity=None)


xgboost_final_model_predictions = xgboost_final_model.predict(X_test_scaled)


print('***** XGBOOST FINAL MODEL *****')                                  #Answer
print('\n')
print(confusion_matrix(y_test, xgboost_final_model_predictions))                           # Metrics
print('\n')
print(classification_report(y_test, xgboost_final_model_predictions))
print('\n')

***** XGBOOST FINAL MODEL *****


[[947 114]
 [147 201]]


              precision    recall  f1-score   support

           0       0.87      0.89      0.88      1061
           1       0.64      0.58      0.61       348

    accuracy                           0.81      1409
   macro avg       0.75      0.74      0.74      1409
weighted avg       0.81      0.81      0.81      1409


print('We got 81% Accuracy with weighted average 81% in Precision, 81% in recall and 81% in F1-score. We see +2 % improvement in all the performance metrics ')

We got 81% Accuracy with weighted average 81% in Precision, 81% in recall and 81% in F1-score. We see +2 % improvement in all the performance metrics


print(xgboost_final_model.score(X_train_scaled, y_train))
print(xgboost_final_model.score(X_test_scaled, y_test))

0.8246361377351793
0.8147622427253371


print('**************************** END OF PART A ***********************************')

**************************** END OF PART A ***********************************


#Importing all the ensemble model classifiers
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import BaggingClassifier
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import AdaBoostClassifier
import xgboost as xgb
from sklearn.pipeline import Pipeline


df_tc1 = pd.read_csv('TelcomCustomer-Churn_1.csv')                      #Answer
df_tc2 = pd.read_csv('TelcomCustomer-Churn_2.csv')


print(df_tc1.shape)
print(df_tc2.shape)

(7043, 10)
(7043, 12)


df2_merged = pd.merge(df_tc1, df_tc2, how= 'outer', on= 'customerID' )     #Answer


df2_merged.to_csv('df_merged.csv')           #Converting dataframe to csv


df2_merged.head()


df2_merged.dtypes

customerID           object
gender               object
SeniorCitizen         int64
Partner              object
Dependents           object
tenure                int64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges         object
Churn                object
dtype: object


def _removing_null_values_(x):
     x = x.replace(r'^\s*$', np.nan, , regex= True)                 #Answer - Replacing the empty strings values with NaN values
#    x.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]
     x['TotalCharges'].replace(np.nan, 2283, inplace= True)            #Imputing average value of 2283 on NaN values


df2_merged['TotalCharges'] = df2_merged['TotalCharges'].apply(_removing_null_values_)


def _datatype_conversion_preprocessing_(x):
    x.astype({'tenure': 'float64',         #Answer
              'SeniorCitizen': 'float64',
              'MonthlyCharges': 'float64',
              'TotalCharges' : 'float64'})


df2_merged['TotalCharges'] = df2_merged['TotalCharges'].apply(_datatype_conversion_preprocessing_)
df2_merged['SeniorCitizen'] = df2_merged['SeniorCitizen'].apply(_datatype_conversion_preprocessing_)
df2_merged['MonthlyCharges'] = df2_merged['MonthlyCharges'].apply(_datatype_conversion_preprocessing_)
df2_merged['TotalCharges'] = df2_merged['TotalCharges'].apply(_datatype_conversion_preprocessing_)


def main_func(x):
     _datatype_conversion_preprocessing_(x)
     _removing_null_values_(x)


main_func(df2_merged)          # Calling main function


df2_merged.dtypes


df2_merged.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]


# OR tradional approach
df2_merged = df2_merged.replace(r'^\s*$', np.nan, regex= True) #Answer - Replacing the empty strings values with NaN values


df2_merged.iloc[[488,753,936,1082,1340,3331,3826,4380,5218,6670,6754]]


df2_merged.isna().sum()  # We replaced null values in TotalCharges column wherever empty strings were there. Total there are 11 records

customerID           0
gender               0
SeniorCitizen        0
Partner              0
Dependents           0
tenure               0
PhoneService         0
MultipleLines        0
InternetService      0
OnlineSecurity       0
OnlineBackup         0
DeviceProtection     0
TechSupport          0
StreamingTV          0
StreamingMovies      0
Contract             0
PaperlessBilling     0
PaymentMethod        0
MonthlyCharges       0
TotalCharges        11
Churn                0
dtype: int64


df2_merged['TotalCharges'].replace(np.nan, 2283, inplace= True) #Answer - Imputing average value of 2283 NaN values


df2_merged = df2_merged.astype({'tenure': 'float64',         #Answer
                              'SeniorCitizen': 'float64',
                              'MonthlyCharges': 'float64',
                              'TotalCharges': 'float64'})


replace_structure_2 = {'gender'          : {'Male' : 0, 'Female' : 1},
                     'Partner'         : {'Yes' : 1, 'No' : 0},
                     'Dependents'      : {'Yes' : 1, 'No' : 0},
                     'PhoneService'    : {'Yes' : 1, 'No' :0},
                     'MultipleLines'   : {'No phone service' : -1, 'Yes' : 1, 'No' : 0},
                     'InternetService' : {'No' : 0, 'DSL' : 1, 'Fiber optic' : 2},
                     'OnlineSecurity'  : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'DeviceProtection' : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'TechSupport'     : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'StreamingTV'     : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'StreamingMovies' : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'Contract'        : {'Month-to-month' : 0, 'One year' : 1, 'Two year' : 2},
                     'OnlineBackup'    : {'No internet service' : -1, 'Yes' : 1, 'No' : 0},
                     'PaperlessBilling': {'Yes' : 1, 'No' :0},
                     'PaymentMethod'   : {'Electronic check' : 0, 'Mailed check' : 1, 'Bank transfer (automatic)' : 2, 'Credit card (automatic)' : 3},
                     'Churn'           : {'Yes' : 1, 'No' :0} }


df2_merged = df2_merged.replace(replace_structure_2)


df2_merged.head()


df2_merged.dtypes

customerID           object
gender               object
SeniorCitizen       float64
Partner              object
Dependents           object
tenure              float64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges        float64
Churn                object
dtype: object


X2 = df2_merged.drop(labels= ['Churn','customerID'], axis=1)    #Separting the data to input and output
y2 = df2_merged['Churn']


X2_train, X2_test, y2_train, y2_test = train_test_split(X2, y2, test_size= 0.20, random_state= 1 )    #Splitting the data into 80:20 proportin where 20% is test data


print(X2_train.shape)
print(X2_test.shape)
print(y2_train.shape)
print(y2_test.shape)

(5634, 19)
(1409, 19)
(5634,)
(1409,)


X2_train.head()


#However, random forest/ensemble methods methods won't have impact with outliers, we will still convert the  features data X as a standard approach, as we have some nummerical features like Total Charges which have high values
X2_train_scaled_2 = X2_train.apply(zscore)
X2_test_scaled_2 = X2_test.apply(zscore)


#Creating different models using pipeline function and instantiating them along with scaling
pl_LogisticRegression = Pipeline( [('Scalar1', StandardScaler()),
                                 ('lr_classifier', LogisticRegression(random_state= 0))] )
pl_DecisionTreeClassifier = Pipeline( [('Scalar2', StandardScaler()),
                                 ('dt_classifier', DecisionTreeClassifier(criterion = 'gini', random_state= 0)) ])
pl_RandomForestClassifier = Pipeline( [('Scalar3', StandardScaler()),
                                 ('rf_classifier', RandomForestClassifier()) ])
pl_BaggingClassifier = Pipeline( [('Scalar4', StandardScaler()),
                                  ('bag_classifier', BaggingClassifier()) ])
pl_GradientBoostingClassifier = Pipeline( [('Scalar5', StandardScaler()),
                                  ('gb_classifier', GradientBoostingClassifier()) ])
pl_AdaBoostClassifier = Pipeline( [('Scalar6', StandardScaler()),
                                  ('ab_classifier', AdaBoostClassifier()) ])


pipe_lines = [pl_LogisticRegression,                      #Creating different ensemble models into a list
              pl_DecisionTreeClassifier,
              pl_RandomForestClassifier,
              pl_BaggingClassifier,
              pl_GradientBoostingClassifier,
              pl_AdaBoostClassifier]


best_accuracy = 0.0                                         #Initially creating the best paramters to 0
best_classifier = 0
best_pipleine = ''


pipe_dict = {0 : 'Logistic Regression',                    #creating a dictionary for all the classifiers for mapping
             1 : 'Decision Tree Classifier',
             2 : 'Random Forest Classifier',
             3 : 'Bagging Classifier',
             4 : 'Gradient Boosting Classifier',
             5 : 'Ada Boost Classifier'}


#This is for just printing all the classifiers
for i, model in enumerate(pipe_lines):
    print(i)
    print(model)

0
Pipeline(steps=[('Scalar1', StandardScaler()),
                ('lr_classifier', LogisticRegression(random_state=0))])
1
Pipeline(steps=[('Scalar2', StandardScaler()),
                ('dt_classifier', DecisionTreeClassifier(random_state=0))])
2
Pipeline(steps=[('Scalar3', StandardScaler()),
                ('rf_classifier', RandomForestClassifier())])
3
Pipeline(steps=[('Scalar4', StandardScaler()),
                ('bag_classifier', BaggingClassifier())])
4
Pipeline(steps=[('Scalar5', StandardScaler()),
                ('gb_classifier', GradientBoostingClassifier())])
5
Pipeline(steps=[('Scalar6', StandardScaler()),
                ('ab_classifier', AdaBoostClassifier())])


# Function to to fit data for each ensemble classifier
for pipe in pipe_lines:
    pipe.fit(X2_train_scaled_2, y2_train)


for i, model in enumerate(pipe_lines):                  #Calculating test accuracy score of test data fro each ensemble classifier
    print(str(pipe_dict[i]) + ' Test Accuracy ' + str(model.score(X2_test_scaled_2, y2_test)) )

Logistic Regression Test Accuracy 0.8076650106458482
Decision Tree Classifier Test Accuracy 0.7366926898509581
Random Forest Classifier Test Accuracy 0.8034066713981547
Bagging Classifier Test Accuracy 0.7963094393186657
Gradient Boosting Classifier Test Accuracy 0.8055358410220014
Ada Boost Classifier Test Accuracy 0.8133427963094393


for i, model in enumerate(pipe_lines):              #Selecting and printing the best classifier based on test accuracy score
     if model.score(X2_test_scaled_2, y2_test) > best_accuracy:
         best_accuracy = model.score(X2_test_scaled_2, y2_test)
         print(best_accuracy)
         best_pipleine = model
         best_classifier = i
        
print('The best model is ' + pipe_dict[best_classifier] + ' with score of ' + str(best_accuracy) )

0.8076650106458482
0.8133427963094393
The best model is Ada Boost Classifier with score of 0.8133427963094393


import pickle


with open('best_model_ab_classifier', 'wb') as files:     # Adding the best model classifier to a pickle file in write mode
    pickle.dump(pl_RandomForestClassifier, files)


with open('best_model_ab_classifier' , 'rb') as files:   # Opening the best model classifier in read mode
    ab = pickle.load(files)


print(ab)

Pipeline(steps=[('Scalar3', StandardScaler()),
                ('rf_classifier', RandomForestClassifier())])


print('**************************** END OF PART B ***********************************')

**************************** END OF PART B ***********************************

	count	mean	std	min	25%	50%	75%	max
SeniorCitizen	7043.0	0.162147	0.368612	0.00	0.0	0.00	0.00	1.00
tenure	7043.0	32.371149	24.559481	0.00	9.0	29.00	55.00	72.00
MonthlyCharges	7043.0	64.761692	30.090047	18.25	35.5	70.35	89.85	118.75

	gender	SeniorCitizen	Partner	Dependents	tenure	PhoneService	MultipleLines	InternetService	OnlineSecurity	OnlineBackup	DeviceProtection	TechSupport	StreamingTV	StreamingMovies	Contract	PaperlessBilling	PaymentMethod	MonthlyCharges	TotalCharges
1814	-0.992573	-0.439475	1.041692	1.543293	-0.825884	0.327916	-0.507510	-1.570293	-1.515358	-1.526412	-1.527759	-1.515762	-1.542669	-1.545217	1.577754	-1.198248	-0.272022	-1.497530	-0.893019
5946	1.007483	-0.439475	-0.959977	-0.647965	0.395961	0.327916	-0.507510	-0.286626	1.318676	1.191604	1.181951	1.309649	-0.221788	1.087704	0.375554	-1.198248	1.469537	0.302996	0.388541
3881	-0.992573	-0.439475	1.041692	-0.647965	1.577078	0.327916	1.046464	-0.286626	1.318676	1.191604	-0.172904	1.309649	-0.221788	-0.228757	1.577754	-1.198248	0.598758	0.012320	1.060276
2389	-0.992573	-0.439475	1.041692	1.543293	1.577078	0.327916	1.046464	-0.286626	1.318676	-0.167404	1.181951	1.309649	1.099093	1.087704	0.375554	-1.198248	-1.142802	0.686687	1.775242
3676	-0.992573	-0.439475	-0.959977	-0.647965	-0.092777	0.327916	-0.507510	-0.286626	1.318676	1.191604	-0.172904	1.309649	1.099093	-0.228757	0.375554	-1.198248	-1.142802	0.186726	-0.104176

	gender	SeniorCitizen	Partner	Dependents	tenure	PhoneService	MultipleLines	InternetService	OnlineSecurity	OnlineBackup	DeviceProtection	TechSupport	StreamingTV	StreamingMovies	Contract	PaperlessBilling	PaymentMethod	MonthlyCharges	TotalCharges
3381	1.017903	-0.441679	-0.993633	-0.678309	0.335987	0.325523	-0.498437	-0.284608	1.311728	-0.198348	1.186837	1.302779	1.101787	1.113014	0.354396	0.810954	0.583597	0.504761	0.453219
6180	1.017903	2.264086	-0.993633	-0.678309	1.353363	0.325523	1.065700	1.000688	1.311728	-0.198348	-0.170512	-0.107064	1.101787	1.113014	1.543310	0.810954	0.583597	1.255673	1.843922
4829	1.017903	-0.441679	-0.993633	-0.678309	-0.844170	0.325523	-0.498437	-0.284608	-0.102369	-0.198348	-0.170512	-0.107064	-0.220545	-0.213751	-0.834518	0.810954	0.583597	-0.655741	-0.780942
3737	1.017903	-0.441679	-0.993633	-0.678309	-1.129035	0.325523	1.065700	-0.284608	-0.102369	-0.198348	-0.170512	-0.107064	-0.220545	-0.213751	-0.834518	0.810954	-0.285629	-0.469262	-0.902046
4249	1.017903	-0.441679	1.006408	1.474254	-0.925560	0.325523	-0.498437	-0.284608	-0.102369	1.138841	1.186837	-0.107064	-0.220545	1.113014	0.354396	0.810954	-0.285629	0.040227	-0.721053

	mean_fit_time	std_fit_time	mean_score_time	std_score_time	param_booster	param_colsample_bytree	param_eta	param_max_depth	param_min_child_weight	params	split0_test_score	split1_test_score	split2_test_score	split3_test_score	mean_test_score	std_test_score	rank_test_score
0	0.338303	0.006762	0.007813	0.007813	gbtree	0.3	0.05	3	1	{'booster': 'gbtree', 'colsample_bytree': 0.3,...	0.801278	0.796309	0.819602	0.780540	0.799432	0.013937	58
1	0.408897	0.077715	0.013644	0.003282	gbtree	0.3	0.05	3	3	{'booster': 'gbtree', 'colsample_bytree': 0.3,...	0.800568	0.797019	0.821733	0.781250	0.800142	0.014431	44
2	0.568119	0.008748	0.012901	0.001724	gbtree	0.3	0.05	3	5	{'booster': 'gbtree', 'colsample_bytree': 0.3,...	0.800568	0.795600	0.820312	0.779119	0.798900	0.014692	74
3	0.695677	0.108269	0.015220	0.000709	gbtree	0.3	0.05	4	1	{'booster': 'gbtree', 'colsample_bytree': 0.3,...	0.801987	0.797019	0.821023	0.780540	0.800142	0.014434	46
4	0.722823	0.130132	0.012810	0.002903	gbtree	0.3	0.05	4	3	{'booster': 'gbtree', 'colsample_bytree': 0.3,...	0.801278	0.799858	0.821023	0.779830	0.800497	0.014573	39

PART A¶

1. A) Read ‘TelcomCustomer-Churn_1.csv’ as a DataFrame and assign it to a variable¶

1. B) Read ‘TelcomCustomer-Churn_2.csv’ as a DataFrame and assign it to a variable¶

1. C) Merge both the DataFrames on key ‘customerID’ to form a single DataFrame¶

1. D) Verify if all the columns are incorporated in the merged DataFrame by using simple comparison Operator in Python¶

2. A) Impute missing/unexpected values in the DataFrame.¶

2. B) Make sure all the variables with continuous values are of ‘Float’ type¶

2. C) Create a function that will accept a DataFrame as input and return pie-charts for all the appropriate Categorical features. Clearly show percentage distribution in the pie-chart¶

2. E) Encode all the appropriate Categorical features with the best suitable approach.¶

2. F) Split the data into 80% train and 20% test¶

2. G) Normalize/Standardize the data with the best suitable approach.¶

3. A) Train a model using XGBoost. Also print best performing parameters along with train and test performance¶

2. B) Improve performance of the XGBoost as much as possible. Also print best performing parameters along with train and test performance.¶

Now we will use the new parameters gamma, eta, colsample_bytree, max_depth & min_child_weight to test the final model metrics - colsample_bytree=0.4, eta=0.1, gamma=0.1, max_depth=4, min_child_weight=5 as the model best test score for these values¶

PART B¶

1. Build a simple ML workflow which will accept a single ‘.csv’ file as input and return a trained base model that can be used for predictions. You can use 1 Dataset from Part 1 (single/merged)¶

2. Create separate functions for various purposes.¶

3. Various base models should be trained to select the best performing model.¶

4. Pickle file should be saved for the best performing model.¶

	customerID	gender	Partner	Dependents	tenure	PhoneService	MultipleLines	InternetService	OnlineSecurity
0	7590-VHVEG	Female	Yes	No	1	No	No phone service	DSL	No
1	5575-GNVDE	Male	No	No	34	Yes	No	DSL	Yes
2	3668-QPYBK	Male	No	No	2	Yes	No	DSL	Yes
3	7795-CFOCW	Male	No	No	45	No	No phone service	DSL	Yes
4	9237-HQITU	Female	No	No	2	Yes	No	Fiber optic	No

	customerID	gender	Partner	Dependents	PhoneService	MultipleLines	InternetService	OnlineSecurity	...	DeviceProtection	TechSupport	StreamingTV	StreamingMovies	Contract	PaperlessBilling	PaymentMethod	MonthlyCharges	Churn
488	4472-LVYGI	Female	Yes	Yes	No	No phone service	DSL	Yes	...	Yes	Yes	Yes	No	Two year	Yes	Bank transfer (automatic)	52.55	No
753	3115-CZMZD	Male	No	Yes	Yes	No	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	Two year	No	Mailed check	20.25	No
936	5709-LVOEQ	Female	Yes	Yes	Yes	No	DSL	Yes	...	Yes	No	Yes	Yes	Two year	No	Mailed check	80.85	No
1082	4367-NUYAO	Male	Yes	Yes	Yes	Yes	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	Two year	No	Mailed check	25.75	No
1340	1371-DWPAZ	Female	Yes	Yes	No	No phone service	DSL	Yes	...	Yes	Yes	Yes	No	Two year	No	Credit card (automatic)	56.05	No
3331	7644-OMVMY	Male	Yes	Yes	Yes	No	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	Two year	No	Mailed check	19.85	No
3826	3213-VVOLG	Male	Yes	Yes	Yes	Yes	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	Two year	No	Mailed check	25.35	No
4380	2520-SGTTA	Female	Yes	Yes	Yes	No	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	Two year	No	Mailed check	20.00	No
5218	2923-ARZLG	Male	Yes	Yes	Yes	No	No	No internet service	...	No internet service	No internet service	No internet service	No internet service	One year	Yes	Mailed check	19.70	No
6670	4075-WKNIU	Female	Yes	Yes	Yes	Yes	DSL	No	...	Yes	Yes	Yes	No	Two year	No	Mailed check	73.35	No
6754	2775-SEFEE	Male	No	Yes	Yes	Yes	DSL	Yes	...	No	Yes	No	No	Two year	Yes	Bank transfer (automatic)	61.90	No

	customerID	gender	Partner	tenure	PhoneService	MultipleLines	InternetService	OnlineSecurity	...	DeviceProtection	TechSupport	Contract	PaperlessBilling	PaymentMethod	MonthlyCharges	TotalCharges	Churn
0	7590-VHVEG	1	1	1.0	0	-1	1	0	...	0	0	0	1	0	29.85	29.85	0
1	5575-GNVDE	0	0	34.0	1	0	1	1	...	1	0	1	0	1	56.95	1889.50	0
2	3668-QPYBK	0	0	2.0	1	0	1	1	...	0	0	0	1	1	53.85	108.15	1
3	7795-CFOCW	0	0	45.0	0	-1	1	1	...	1	1	1	0	2	42.30	1840.75	0
4	9237-HQITU	1	0	2.0	1	0	2	0	...	0	0	0	1	0	70.70	151.65	1

	gender	Partner	Dependents	tenure	PhoneService	MultipleLines	InternetService	OnlineSecurity	OnlineBackup	DeviceProtection	TechSupport	StreamingTV	StreamingMovies	Contract	PaymentMethod	MonthlyCharges	TotalCharges
1814	0	1	1	12.0	1	0	0	-1	-1	-1	-1	-1	-1	2	1	19.70	258.35
5946	1	0	0	42.0	1	0	1	1	1	1	1	0	1	1	3	73.90	3160.55
3881	0	1	0	71.0	1	1	1	1	1	0	1	0	0	2	2	65.15	4681.75
2389	0	1	1	71.0	1	1	1	1	0	1	1	1	1	1	0	85.45	6300.85
3676	0	0	0	30.0	1	0	1	1	1	0	1	1	0	1	0	70.40	2044.75

PART A¶

1. A) Read ‘TelcomCustomer-Churn_1.csv’ as a DataFrame and assign it to a variable¶

1. B) Read ‘TelcomCustomer-Churn_2.csv’ as a DataFrame and assign it to a variable¶

1. C) Merge both the DataFrames on key ‘customerID’ to form a single DataFrame¶

1. D) Verify if all the columns are incorporated in the merged DataFrame by using simple comparison Operator in Python¶

2. A) Impute missing/unexpected values in the DataFrame.¶

2. B) Make sure all the variables with continuous values are of ‘Float’ type¶

2. C) Create a function that will accept a DataFrame as input and return pie-charts for all the appropriate Categorical features. Clearly show percentage distribution in the pie-chart¶

2. D) Share insights for Q2.c.¶

2. E) Encode all the appropriate Categorical features with the best suitable approach.¶

2. F) Split the data into 80% train and 20% test¶

2. G) Normalize/Standardize the data with the best suitable approach.¶

3. A) Train a model using XGBoost. Also print best performing parameters along with train and test performance¶

2. B) Improve performance of the XGBoost as much as possible. Also print best performing parameters along with train and test performance.¶

Now we will use the new parameters gamma, eta, colsample_bytree, max_depth & min_child_weight to test the final model metrics - colsample_bytree=0.4, eta=0.1, gamma=0.1, max_depth=4, min_child_weight=5 as the model best test score for these values¶

PART B¶

1. Build a simple ML workflow which will accept a single ‘.csv’ file as input and return a trained base model that can be used for predictions. You can use 1 Dataset from Part 1 (single/merged)¶

2. Create separate functions for various purposes.¶

3. Various base models should be trained to select the best performing model.¶

4. Pickle file should be saved for the best performing model.¶